Summary of Pattern Comparison and Concluding Remarks

366 R Grayson and G Blo¨ schl
increase as hydrologists come to realise the value of patterns for model development
and testing. Comprehensive uncertainty analysis also needs to be further
developed for spatially distributed models. The potential of these methods for
assessing the separate sources of uncertainty (input information, parameter
values, model structure and data used in testing) is large, but at present they
are not computationally tractable for most distributed models. Methodological
developments as well as improvements in computer power are likely to lead to
wider use of such techniques.
14.3.3 Challenges for Model Conceptualisation
The primary challenge of hydrologists has been, and remains, the prediction
of hydrological response in ‘‘ungauged’’ areas – i.e. areas for which we have no
hydrological response information. There is still a need to improve methods for
generalising results from small catchments such as those described in this book,
to other catchments; from small catchments to large catchments; and for being
able to predict hydrological response under changed land use and climatic conditions
in catchments of all sizes. All of these needs can be met only with better
understanding and representation of fundamental processes, and their spatial
variability across a range of scales. Distributed modelling generally has moved
beyond just trying to scale up small catchment models to large scales because of
problems with identifiability and scale dependence. As has been suggested for
some time by many authors, we need models for a range of scales that are
parsimonious, but that reflect the manifestation of important processes at
those different scales. In moving beyond the notion of ‘‘trying to model everything’’
we should be developing methods to identify dominant processes that
control hydrological response in different environments (landscapes and climates)
and at different scales, and then develop models to focus on these dominant
processes (a notion we might call the ‘‘Dominant Processes Concept’’
(DPC)). This would provide a framework for the development and application
of techniques specially designed to deal with those controls and help to avoid
some of the overparameterisation problems that occur when processes that are
not important are represented in models. Developments along the lines of the
DPC may help with the generalisation problems that have haunted hydrologists
since the science began.
14.4 FINAL REMARKS
As mentioned in the introduction, there have been many calls for data collection
and analysis to go hand in hand, for improved understanding of processes,
and for the scientific endeavour of measurement to be recognised. There is
a range of evidence that these calls have elicited a response. For example,
Water Resources Research has had ‘‘data notes’’ for some time (Hornberger,
1994) and the number being published is increasing. There is an increasing
awareness that the development of a spatial model is not of itself useful, unless